Motion and Vision Combine to Detect Flaws
Plastic film is highly valued in the computer industry where plastic sheeting is used in the creation of printed circuit boards (PCBs). MicroCraft, a Japanese company, was challenged by a plastic sheeting manufacturer to come up with an inspection system that could find and categorize the micron-sized defects that can occur in plastic film production.
The manufacturer’s current method for detecting defects required an operator to evaluate an image projected onto a wall with an overhead projector. The operator would circle defects on the sheet and then carry the sheet to a microscope with a film camera attached (to photograph the defect). Next, the operator estimated the defect
|National Instruments’ PXI IMAQ 1422 board.|
area, length, and width based on other calibrated pictures and made a handwritten log entry for all the information. The operator then placed the picture in a “picture book,” in a section for the defects category. This process was time-consuming and only done sporadically, because no more than 10 samples could be completed in an 8-hour work shift.
Defects appear as bright spots
MicroCraft re-conducted the manufacturer’s original test procedure using a cross-polarized light field, along with vision and motion systems. In this method, distortion of the aligned polymer chains in the material made defects appear as bright spots. A two-step analysis was then conducted using two cameras—ones with large and small fields of view (LFOV and SFOV, respectively) and different magnifications (LFOV: 0.3X, SFOV: 6.4X). The LFOV camera was used for initial film inspection to find defects. Operators used the SFOV camera for secondary defection evaluation, and to capture images and measurements for data storage.
|Other articles in the May 2009 Control Engineering North American print edition supplement:
– 415 Parts Seen
– Intelligent Vision Stops Bypass of Quality Control
– The New Look of Facial Recognition
A backlight table was covered with a polarizing sheet and placed on a 2-axis linear rail system to control the lighting and motion of the manufacturer’s 8-in. diameter samples. Motion, vision, and data collection were combined using a National Instruments PXI-7324 motion board and two PXI IMAQ boards.
After the operator orients a sample for correct polarization, the system automatically scans its 16 2-in square “quadrants.” Using the LFOV camera, the operator looks for defects. Using vision analysis virtual instruments (VIs) to determine the center of a suspected defect, the PXI system commands the table to move so that each potential defect is in the field of view for high magnification. Several image analysis VIs also are used to counter common lens problems, such as pin-cushioning. Motion analysis VIs create an algorithm that properly positions the defect area directly under the SFOV camera.
When the operator locates a suspect area using the SFOV camera, a live image appears. The operator can reject the image—if it is a dust speck, for example—or classify the defect and make calibrated measurements for area, length, and width. Resolution in the SFOV camera was achieved through calculations using NI Vision Builder and a USAF resolution target at 1.17 microns/pixel. With this resolution, the system can detect and quantify defects in the manufacturer’s required 10-100 micron range. At roughly 15 minutes per sample, operators can now test up to 30 samples per shift.